Export legacy and contact-sheet captions from canonical SQLite

This commit is contained in:
Sleeps
2026-10-08 11:03:16 +01:00
parent 895c72465f
commit cf44aa686c
+49 -20
View File
@@ -1,26 +1,55 @@
docker compose exec -T scene-search python - <<'PY' > captions.txt #!/usr/bin/env bash
set -Eeuo pipefail
source "$(cd -- "$(dirname -- "$0")" && pwd)/common.sh"
need docker
output_file="$REPO_ROOT/captions.txt"
temp_file="$(mktemp "$REPO_ROOT/.captions.txt.XXXXXX")"
trap 'rm -f -- "$temp_file"' EXIT
# Export from the canonical SQLite database. The newer contact-sheet captions
# live in caption_windows, while older per-frame captions live in frames.
# Write to a temporary file so a Docker/SQL failure cannot truncate a
# previously successful export.
compose exec -T scene-search python - <<'PY' > "$temp_file"
import sqlite3 import sqlite3
con = sqlite3.connect('/app/data/scenes.sqlite3') con = sqlite3.connect("/app/data/scenes.sqlite3")
con.row_factory = sqlite3.Row con.row_factory = sqlite3.Row
try:
rows = con.execute("""
SELECT m.title, m.jellyfin_item_id,
cw.start_ms AS timestamp_ms, cw.caption
FROM caption_windows cw
JOIN media m ON m.id = cw.media_id
WHERE cw.caption IS NOT NULL AND TRIM(cw.caption) != ''
for row in con.execute(""" UNION ALL
SELECT
m.title,
m.jellyfin_item_id,
f.timestamp_ms,
f.caption
FROM media m
JOIN frames f ON f.media_id = m.id
WHERE m.source_kind = 'jellyfin'
AND f.caption IS NOT NULL
AND TRIM(f.caption) != ''
ORDER BY m.title, f.timestamp_ms
"""):
seconds = row['timestamp_ms'] // 1000
mins, secs = divmod(seconds, 60)
hours, mins = divmod(mins, 60)
print(f"\n{row['title']} ({row['jellyfin_item_id']}) [{hours:02d}:{mins:02d}:{secs:02d}]") SELECT m.title, m.jellyfin_item_id,
print(row['caption']) f.timestamp_ms, f.caption
FROM frames f
JOIN media m ON m.id = f.media_id
WHERE f.caption IS NOT NULL AND TRIM(f.caption) != ''
ORDER BY title, timestamp_ms
""")
for row in rows:
seconds = int(row["timestamp_ms"]) // 1000
hours, remainder = divmod(seconds, 3600)
minutes, seconds = divmod(remainder, 60)
timestamp = f"{hours:02d}:{minutes:02d}:{seconds:02d}"
caption = " ".join(row["caption"].split())
print(f"{row['title']} [{timestamp}]")
print(f"Jellyfin ID: {row['jellyfin_item_id']}")
print(f"Caption: {caption}")
print()
finally:
con.close()
PY PY
mv -f -- "$temp_file" "$output_file"
printf 'Saved all captions to %s\n' "$output_file"